Many automatic translation works have been addressed between major European language pairs, by taking advantage of large scale parallel corpora, but very few research works are conducted on the Amharic-Arabic language pair due to its parallel data scarcity. Two Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU) based Neural Machine Translation (NMT) models are developed using Attention-based Encoder-Decoder architecture which is adapted from the open-source OpenNMT system. In order to perform the experiment, a small parallel Quranic text corpus is constructed by modifying the existing monolingual Arabic text and its equivalent translation of Amharic language text corpora available on Tanzile. LSTM and GRU based NMT models and Google Translation system are compared and found that LSTM based OpenNMT outperforms GRU based OpenNMT and Google Translation system, with a BLEU score of 12%, 11%, and 6% respectively.
Sequential word order is important when processing text. Currently, neural networks (NNs) address this by modeling word position using position embeddings. The problem is that position embeddings capture the position of individual words, but not the ordered relationship (e.g., adjacency or precedence) between individual word positions. We present a novel and principled solution for modeling both the global absolute positions of words and their order relationships. Our solution generalizes word embeddings, previously defined as independent vectors, to continuous word functions over a variable (position). The benefit of continuous functions over variable positions is that word representations shift smoothly with increasing positions. Hence, word representations in different positions can correlate with each other in a continuous function. The general solution of these functions is extended to complex-valued domain due to richer representations. We extend CNN, RNN and Transformer NNs to complex-valued versions to incorporate our complex embedding (we make all code available). Experiments on text classification, machine translation and language modeling show gains over both classical word embeddings and position-enriched word embeddings. To our knowledge, this is the first work in NLP to link imaginary numbers in complex-valued representations to concrete meanings (i.e., word order).
To transcribe spoken language to written medium, most alphabets enable an unambiguous sound-to-letter rule. However, some writing systems have distanced themselves from this simple concept and little work exists in Natural Language Processing (NLP) on measuring such distance. In this study, we use an Artificial Neural Network (ANN) model to evaluate the transparency between written words and their pronunciation, hence its name Orthographic Transparency Estimation with an ANN (OTEANN). Based on datasets derived from Wikimedia dictionaries, we trained and tested this model to score the percentage of correct predictions in phoneme-to-grapheme and grapheme-to-phoneme translation tasks. The scores obtained on 17 orthographies were in line with the estimations of other studies. Interestingly, the model also provided insight into typical mistakes made by learners who only consider the phonemic rule in reading and writing.
Self-attention based Transformer has demonstrated the state-of-the-art performances in a number of natural language processing tasks. Self-attention is able to model long-term dependencies, but it may suffer from the extraction of irrelevant information in the context. To tackle the problem, we propose a novel model called \textbf{Explicit Sparse Transformer}. Explicit Sparse Transformer is able to improve the concentration of attention on the global context through an explicit selection of the most relevant segments. Extensive experimental results on a series of natural language processing and computer vision tasks, including neural machine translation, image captioning, and language modeling, all demonstrate the advantages of Explicit Sparse Transformer in model performance. We also show that our proposed sparse attention method achieves comparable or better results than the previous sparse attention method, but significantly reduces training and testing time. For example, the inference speed is twice that of sparsemax in Transformer model. Code will be available at \url{https://github.com/lancopku/Explicit-Sparse-Transformer}
Multilingual neural machine translation (NMT) has recently been investigated from different aspects (e.g., pivot translation, zero-shot translation, fine-tuning, or training from scratch) and in different settings (e.g., rich resource and low resource, one-to-many, and many-to-one translation). This paper concentrates on a deep understanding of multilingual NMT and conducts a comprehensive study on a multilingual dataset with more than 20 languages. Our results show that (1) low-resource language pairs benefit much from multilingual training, while rich-resource language pairs may get hurt under limited model capacity and training with similar languages benefits more than dissimilar languages; (2) fine-tuning performs better than training from scratch in the one-to-many setting while training from scratch performs better in the many-to-one setting; (3) the bottom layers of the encoder and top layers of the decoder capture more language-specific information, and just fine-tuning these parts can achieve good accuracy for low-resource language pairs; (4) direct translation is better than pivot translation when the source language is similar to the target language (e.g., in the same language branch), even when the size of direct training data is much smaller; (5) given a fixed training data budget, it is better to introduce more languages into multilingual training for zero-shot translation.
Multi-modal machine translation aims at translating the source sentence into a different language in the presence of the paired image. Previous work suggests that additional visual information only provides dispensable help to translation, which is needed in several very special cases such as translating ambiguous words. To make better use of visual information, this work presents visual agreement regularized training. The proposed approach jointly trains the source-to-target and target-to-source translation models and encourages them to share the same focus on the visual information when generating semantically equivalent visual words (e.g. "ball" in English and "ballon" in French). Besides, a simple yet effective multi-head co-attention model is also introduced to capture interactions between visual and textual features. The results show that our approaches can outperform competitive baselines by a large margin on the Multi30k dataset. Further analysis demonstrates that the proposed regularized training can effectively improve the agreement of attention on the image, leading to better use of visual information.
State-of-the-art Transformer-based neural machine translation (NMT) systems still follow a standard encoder-decoder framework, in which source sentence representation can be well done by an encoder with self-attention mechanism. Though Transformer-based encoder may effectively capture general information in its resulting source sentence representation, the backbone information, which stands for the gist of a sentence, is not specifically focused on. In this paper, we propose an explicit sentence compression method to enhance the source sentence representation for NMT. In practice, an explicit sentence compression goal used to learn the backbone information in a sentence. We propose three ways, including backbone source-side fusion, target-side fusion, and both-side fusion, to integrate the compressed sentence into NMT. Our empirical tests on the WMT English-to-French and English-to-German translation tasks show that the proposed sentence compression method significantly improves the translation performances over strong baselines.
Lectures translation is a case of spoken language translation and there is a lack of publicly available parallel corpora for this purpose. To address this, we examine a language independent framework for parallel corpus mining which is a quick and effective way to mine a parallel corpus from publicly available lectures at Coursera. Our approach determines sentence alignments, relying on machine translation and cosine similarity over continuous-space sentence representations. We also show how to use the resulting corpora in a multistage fine-tuning based domain adaptation for high-quality lectures translation. For Japanese--English lectures translation, we extracted parallel data of approximately 40,000 lines and created development and test sets through manual filtering for benchmarking translation performance. We demonstrate that the mined corpus greatly enhances the quality of translation when used in conjunction with out-of-domain parallel corpora via multistage training. This paper also suggests some guidelines to gather and clean corpora, mine parallel sentences, address noise in the mined data, and create high-quality evaluation splits. For the sake of reproducibility, we will release our code for parallel data creation.
An effective method to generate a large number of parallel sentences for training improved neural machine translation (NMT) systems is the use of the back-translations of the target-side monolingual data. The standard back-translation method has been shown to be unable to efficiently utilize the available huge amount of existing monolingual data because of the inability of translation models to differentiate between the authentic and synthetic parallel data during training. Tagging, or using gates, has been used to enable translation models to distinguish between synthetic and authentic data, improving standard back-translation and also enabling the use of iterative back-translation on language pairs that underperformed using standard back-translation. In this work, we approach back-translation as a domain adaptation problem, eliminating the need for explicit tagging. In the approach -- \emph{tag-less back-translation} -- the synthetic and authentic parallel data are treated as out-of-domain and in-domain data respectively and, through pre-training and fine-tuning, the translation model is shown to be able to learn more efficiently from them during training. Experimental results have shown that the approach outperforms the standard and tagged back-translation approaches on low resource English-Vietnamese and English-German neural machine translation.
Word translation is a problem in machine translation that seeks to build models that recover word level correspondence between languages. Recent approaches to this problem have shown that word translation models can learned with very small seeding dictionaries, and even without any starting supervision. In this paper we propose a method to jointly find translations between a pair of languages. Not only does our method learn translations in both directions but it improves accuracy of those translations over past methods.
Learning word embeddings using distributional information is a task that has been studied by many researchers, and a lot of studies are reported in the literature. On the contrary, less studies were done for the case of multiple languages. The idea is to focus on a single representation for a pair of languages such that semantically similar words are closer to one another in the induced representation irrespective of the language. In this way, when data are missing for a particular language, classifiers from another language can be used.
Machine translation (MT) is an important task in natural language processing (NLP) as it automates the translation process and reduces the reliance on human translators. With the resurgence of neural networks, the translation quality surpasses that of the translations obtained using statistical techniques for most language-pairs. Up until a few years ago, almost all of the neural translation models translated sentences independently, without incorporating the wider document-context and inter-dependencies among the sentences. The aim of this survey paper is to highlight the major works that have been undertaken in the space of document-level machine translation after the neural revolution, so that researchers can recognise the current state and future directions of this field. We provide an organisation of the literature based on novelties in modelling and architectures as well as training and decoding strategies. In addition, we cover evaluation strategies that have been introduced to account for the improvements in document MT, including automatic metrics and discourse-targeted test sets. We conclude by presenting possible avenues for future exploration in this research field.
Transformer-based architectures represent the state of the art in sequence modeling tasks like machine translation and language understanding. Their applicability to multi-modal contexts like image captioning, however, is still largely under-explored. With the aim of filling this gap, we present M$^2$ - a Meshed Transformer with Memory for Image Captioning. The architecture improves both the image encoding and the language generation steps: it learns a multi-level representation of the relationships between image regions integrating learned a priori knowledge, and uses a mesh-like connectivity at decoding stage to exploit low- and high-level features. Experimentally, we investigate the performance of the M$^2$ Transformer and different fully-attentive models in comparison with recurrent ones. When tested on COCO, our proposal achieves a new state of the art in single-model and ensemble configurations on the "Karpathy" test split and on the online test server. We also assess its performances when describing objects unseen in the training set. Trained models and code for reproducing the experiments are publicly available at: https://github.com/aimagelab/meshed-memory-transformer.
Previous studies on the domain adaptation for neural machine translation (NMT) mainly focus on the one-pass transferring out-of-domain translation knowledge to in-domain NMT model. In this paper, we argue that such a strategy fails to fully extract the domain-shared translation knowledge, and repeatedly utilizing corpora of different domains can lead to better distillation of domain-shared translation knowledge. To this end, we propose an iterative dual domain adaptation framework for NMT. Specifically, we first pre-train in-domain and out-of-domain NMT models using their own training corpora respectively, and then iteratively perform bidirectional translation knowledge transfer (from in-domain to out-of-domain and then vice versa) based on knowledge distillation until the in-domain NMT model convergences. Furthermore, we extend the proposed framework to the scenario of multiple out-of-domain training corpora, where the above-mentioned transfer is performed sequentially between the in-domain and each out-of-domain NMT models in the ascending order of their domain similarities. Empirical results on Chinese-English and English-German translation tasks demonstrate the effectiveness of our framework.
Current approaches to machine translation (MT) either translate sentences in isolation, disregarding the context they appear in, or model context at the level of the full document, without a notion of any internal structure the document may have. In this work we consider the fact that documents are rarely homogeneous blocks of text, but rather consist of parts covering different topics. Some documents, such as biographies and encyclopedia entries, have highly predictable, regular structures in which sections are characterised by different topics. We draw inspiration from Louis and Webber (2014) who use this information to improve statistical MT and transfer their proposal into the framework of neural MT. We compare two different methods of including information about the topic of the section within which each sentence is found: one using side constraints and the other using a cache-based model. We create and release the data on which we run our experiments - parallel corpora for three language pairs (Chinese-English, French-English, Bulgarian-English) from Wikipedia biographies, which we extract automatically, preserving the boundaries of sections within the articles.
Future human action forecasting from partial observations of activities is an important problem in many practical applications such as assistive robotics, video surveillance and security. We present a method to forecast actions for the unseen future of the video using a neural machine translation technique that uses encoder-decoder architecture. The input to this model is the observed RGB video, and the objective is to forecast the correct future symbolic action sequence. Unlike prior methods that make action predictions for some unseen percentage of video one for each frame, we predict the complete action sequence that is required to accomplish the activity. We coin this task action sequence forecasting. To cater for two types of uncertainty in the future predictions, we propose a novel loss function. We show a combination of optimal transport and future uncertainty losses help to improve results. We extend our action sequence forecasting model to perform weakly supervised action forecasting on two challenging datasets, the Breakfast and the 50Salads. Specifically, we propose a model to predict actions of future unseen frames without using frame level annotations during training. Using Fisher vector features, our supervised model outperforms the state-of-the-art action forecasting model by 0.83% and 7.09% on the Breakfast and the 50Salads datasets respectively. Our weakly supervised model is only 0.6% behind the most recent state-of-the-art supervised model and obtains comparable results to other published fully supervised methods, and sometimes even outperforms them on the Breakfast dataset. Most interestingly, our weakly supervised model outperforms prior models by 1.04% leveraging on proposed weakly supervised architecture, and effective use of attention mechanism and loss functions.
Lexical ambiguity, a challenging phenomenon in all natural languages, is particularly prevalent for languages with diacritics that tend to be omitted in writing, such as Arabic. Omitting diacritics leads to an increase in the number of homographs: different words with the same spelling. Diacritic restoration could theoretically help disambiguate these words, but in practice, the increase in overall sparsity leads to performance degradation in NLP applications. In this paper, we propose approaches for automatically marking a subset of words for diacritic restoration, which leads to selective homograph disambiguation. Compared to full or no diacritic restoration, these approaches yield selectively-diacritized datasets that balance sparsity and lexical disambiguation. We evaluate the various selection strategies extrinsically on several downstream applications: neural machine translation, part-of-speech tagging, and semantic textual similarity. Our experiments on Arabic show promising results, where our devised strategies on selective diacritization lead to a more balanced and consistent performance in downstream applications.
As a special machine translation task, dialect translation has two main characteristics: 1) lack of parallel training corpus; and 2) possessing similar grammar between two sides of the translation. In this paper, we investigate how to exploit the commonality and diversity between dialects thus to build unsupervised translation models merely accessing to monolingual data. Specifically, we leverage pivot-private embedding, layer coordination, as well as parameter sharing to sufficiently model commonality and diversity among source and target, ranging from lexical, through syntactic, to semantic levels. In order to examine the effectiveness of the proposed models, we collect 20 million monolingual corpus for each of Mandarin and Cantonese, which are official language and the most widely used dialect in China. Experimental results reveal that our methods outperform rule-based simplified and traditional Chinese conversion and conventional unsupervised translation models over 12 BLEU scores.
Sequence-level knowledge distillation (SLKD) is a model compression technique that leverages large, accurate teacher models to train smaller, under-parameterized student models. Why does pre-processing MT data with SLKD help us train smaller models? We test the common hypothesis that SLKD addresses a capacity deficiency in students by "simplifying" noisy data points and find it unlikely in our case. Models trained on concatenations of original and "simplified" datasets generalize just as well as baseline SLKD. We then propose an alternative hypothesis under the lens of data augmentation and regularization. We try various augmentation strategies and observe that dropout regularization can become unnecessary. Our methods achieve BLEU gains of 0.7-1.2 on TED Talks.